MiniMax Open-Sources M2.7, a 230B Parameter Model That Tops SWE-Pro and Terminal Bench
Chinese AI lab MiniMax has released its M2.7 model as fully open source, posting state-of-the-art results on software engineering benchmarks — and it already runs locally.
MiniMax, the Chinese AI lab that has been steadily climbing the performance charts, officially open-sourced its M2.7 model over the weekend, posting state-of-the-art scores on two of the most demanding software engineering benchmarks in the field. As @MiniMax_AI announced, M2.7 achieved 56.22% on SWE-Pro and 57.0% on Terminal Bench 2 — numbers that place it ahead of several proprietary models on tasks that require real-world code understanding and execution.
Unlock the full briefing
Get every story in today's briefing, the full archive, and the daily AI intelligence brief.
All stories today
Full archive
Daily brief
Cancel anytime. Payments powered by Stripe.